Automated Mineralogy As a Tool for Geolocation Based on Soil Trace

ORE Open Research Exeter

TITLE Automated mineralogical proﬁling of soils as an indicator of local bedrock lithology: a tool for predictive forensic geolocation

AUTHORS Pirrie, D; Crean, DE; Pidduck, AJ; et al.

JOURNAL Geological Society, London, Special Publications

DEPOSITED IN ORE 11 December 2019

This version available at

http://hdl.handle.net/10871/40081

Open Research Exeter makes this work available in accordance with publisher policies.

A NOTE ON VERSIONS

The version presented here may diﬀer from the published version. If citing, you are advised to consult the published version for pagination, volume/issue and date of publication OFFICIAL

1 Automated mineralogical profiling of soils as an indicator of local 2 bedrock lithology: a tool for predictive forensic geolocation 3 4 Duncan Pirriea,d, Daniel E. Creanb, Allan J. Pidduckb, Timothy M. Nichollsb, 5 Roy P. Awberyb, Robin K. Shailc

8 aHelford Geoscience LLP, Trelowarren Mill Barn, Mawgan, Helston, Cornwall,

9 TR12 6AE, UK 10 11 bAWE Plc, Aldermaston, Reading, RG7 4PR, UK 12 13 cCamborne School of Mines, University of Exeter, Penryn Campus, Penryn, 14 Cornwall, TR10 9FE, UK 15 16 dSchool of Applied Sciences, Faculty of Computing, Engineering and

17 Sciences, University of South Wales, Glyntaf, Pontypridd, Rhondda Cynon 18 Taff, CF37 4AT, UK 19 20 ABSTRACT 21 22 The use of soil evidence to identify an unknown location is a powerful tool to 23 determine the provenance of an item in an investigation. We are particularly 24 interested in the use of these indicators in nuclear forensic cases, whereby

25 identification of locations associated with for example, a smuggled nuclear 26 material, may be used to indicate the provenance of a find. The use of soil 27 evidence to identify an unknown location relies on understanding and 28 predicting how soils vary in composition depending on their geological / 29 geographical setting. In this study, compositional links between the 30 mineralogy of forty soils and the underlying bedrock geology were 31 established. The soil samples were collected from locations with broadly 32 similar climate and land use across a range of geological settings in a ‘test 33 bed’ 3500 km2 area of South West England. In this region, the soils formed 34 through chemical weathering of the bedrock, representing a worst case for © British Crown Owned Copyright 2018/AWE 1

OFFICIAL

35 this type of forensic geolocation due to the high degree of alteration of the 36 parent rock during soil formation. The mineralogy was quantified using 37 automated SEM-EDX analysis. The soil mineralogy and texture are consistent 38 with the underlying geology as indicated by regional-scale geological 39 mapping. 40 41 Keywords: soil forensics, provenancing. 42

43 Introduction 44 45 It is widely recognised that soil is a useful class of trace evidence which 46 can be used to test a possible association between a geographical location 47 and material recovered from, for example, footwear, clothing, vehicles and 48 objects used during an offence (e.g. Bull et al., 2006; Pirrie and Ruffell, 2012). 49 In general, most published forensic soil science studies have addressed the 50 use of soil as an evidential tool, testing an association between an object and

51 a known location through comparative analysis (Fitzpatrick et al., 2009). 52 However, soil analysis can also be used to attempt to identify unknown 53 locations based on the soil characteristics (Bowen and Craven, 2013; Lark 54 and Rawlins, 2008; Owens et al., 2016). This use of soil trace evidence to 55 determine the geographic position of otherwise unknown locations can be 56 referred to as predictive geolocation (Pirrie et al., 2017) and is of increasing 57 interest to the forensic geoscience community (e.g. Stern et al., 2019; Caritat 58 et al., 2019).

59 One aspect of our interest in predictive geolocation relates to its 60 potential to determine the geographic source (provenance) of an item in a 61 nuclear forensic investigation (Mayer et al., 2012). These scenarios involve 62 the discovery or interdiction of nuclear material outside regulatory or 63 institutional control, such as so-called ‘nuclear smuggling’ cases (Wallenius et 64 al., 2006; Mayer et al., 2015). In these investigations, there may be a wide 65 range of potential sources of the material, (e.g. facilities which have 66 processed the particular type of material), and therefore the potential to 67 constrain the possible origins of a find through predictive geolocation analysis 68 would be of great value in establishing the provenance of a found material. © British Crown Owned Copyright 2018/AWE 2

OFFICIAL

69 This is complementary to the types of analysis which fall under the field of 70 nuclear forensic science (Keegan et al., 2016), which relate to the properties 71 of the material itself, such as the identification of particular processes through 72 the evaluation of signature chemical properties of the material. 73 Predictive geolocation analysis relies on signatures accumulated from 74 the environments to which a nuclear material has been exposed during 75 production, transport and storage; both before and after institutional control 76 was lost. Packing and repackaging of materials may introduce a number of

77 internal interfaces where materials from different environments may 78 accumulate. These environmental signatures may include natural materials 79 such as soils and dusts, to which forensic soil geolocation analysis techniques 80 can be applied to indicate location(s) associated with the sample. 81 For this predictive geolocation analysis to be practical, it is necessary 82 to select key parameters which can be measured in a realistic (short) 83 timescale for use in an investigation, and which have a high degree of 84 specificity in indicating an unknown location. Owing to the wide availability of

85 spatial geological reference datasets (e.g. from national geological surveys), 86 indicators related to the underlying lithology at a location are useful in this 87 context (Pirrie et al., 2013). Determination of the likely source geology from 88 soil analysis would allow an investigation to be focussed to areas where this 89 setting is known to occur. For this approach to have forensic validity, the 90 relationship between soil mineralogy and underlying geology has to be 91 established. Soil-forming processes can profoundly alter the composition of 92 the parent material depending on environmental conditions, and therefore

93 there is a need to understand the specificity with which lithological signatures 94 could be detected in a range of soil forming environments. 95 In this paper we present a systematic study of the relationship between 96 the underlying geology and soil mineralogy in a geologically varied 3500 km2 97 area of SW England. The area provides a ‘test bed’ with a variety of well 98 documented major rock types and settings. Soils within the area are mainly 99 mature loams, developed through chemical weathering and human 100 management. This represents an advanced state of alteration of the 101 underlying rock and therefore a high potential for disturbance of the 102 mineralogical signatures linking the soil to the associated bedrock. Automated © British Crown Owned Copyright 2018/AWE 3

OFFICIAL

103 mineralogy based on scanning electron microscopy was used as a rapid 104 method for characterisation the mineralogy and texture of the soils. This tests 105 the link between rapidly measureable soil characteristics (in this case 106 mineralogy) and widely available geological datasets, such that an indication 107 of potential geological settings could be provided from an unknown soil 108 sample in an investigation. 109 110 Overview of Study Area

111 112 In this study 35 geographical locations over an area of approximately 113 3500 km2 in Cornwall and Devon, SW England, UK were selected for soil 114 sampling (Fig. 1, Table 1). Factors such as variations in climate and land-use 115 were minimised as far as possible, such that the dataset allows the 116 significance of the underlying bedrock geology as the main control on soil 117 mineralogy to be tested. All of the sampling locations would also have had 118 surface superficial sediments present mantling the bedrock geology to

119 variable extents. 120 The selected sampling area whilst geologically varied, shows the 121 repetition of a number of major rock types, and samples from these were 122 taken to test discrimination between soils developed on different age units 123 composed of the same dominant rock types. SW England has also been a 124 recent focus for baseline environmental and geochemical surveys based on 125 both airborne datasets and regional surface soil and sediment sampling 126 programmes (Beamish, 2015; Kirkwood et al., 2016).

127 128 Geology

129 130 SW England is an Upper Palaeozoic massif whose bedrock geology, in 131 common with much of western Europe, was strongly influenced by the late 132 Palaeozoic Variscan mountain-building episode (Shail and Leveridge, 2009). 133 The near-surface geology is dominated by Devonian and Carboniferous 134 metasedimentary rocks (Fig 1.). Other significant features include associated 135 minor basic intrusive / extrusive igneous rocks, a major granite batholith which 136 was emplaced in the Early Permian, and the Lizard Complex, which includes © British Crown Owned Copyright 2018/AWE 4

OFFICIAL

137 partially serpentinised mantle peridotites, gabbros and high-grade 138 metamorphosed mafic igneous rocks. Devonian and Carboniferous 139 successions occur in six east-west trending sedimentary basins (Leveridge, 140 2011; Leveridge and Shail, 2011a, b). These formed during rifting which 141 resulted in the formation of a passive margin and associated oceanic 142 lithosphere. Sedimentation in the Gramscatho and Culm basins continued 143 during the early stages of the Variscan continental collision (Leveridge, 2011; 144 Leveridge and Shail, 2011a). The Looe Basin includes non-marine

145 sandstones and mudstones (Leveridge, 2011) but, more generally, the basin 146 successions comprise mudstones and sandstones, along with minor 147 limestones and cherts, deposited in shallow to deep-marine environments 148 (Leveridge and Shail, 2011b). Rift-related mafic igneous rocks (basalts, 149 dolerites and gabbros) are locally important constituents of the basin fills and 150 sedimentary exhalative (SedEx) and volcanic massive sulphide (VMS) 151 mineralisation styles locally occur (Benham et al., 2005). The Lizard Complex 152 includes partially serpentinised mantle peridotites, gabbros and high-grade

153 metamorphosed mafic igneous rocks; it represents a fragment of the oceanic 154 lithosphere (an ophiolite) formed during rifting and also includes small areas of 155 high-grade continental metamorphic rocks. 156 Variscan continental collision brought about deformation and low-grade 157 regional metamorphism of all of the basinal successions (Shail and Leveridge, 158 2009). Consequently the rocks are variably thrust-faulted and folded and one 159 or more cleavages are developed. Quartz veins, precipitated from 160 metamorphic fluids, are ubiquitous and locally associated with precious metal

161 mineralisation. Continental collision continued until the latest Carboniferous, 162 when it was followed by a NNW-SSE extensional regime which persisted 163 through most of the Early Permian. This extension led to the development of 164 post-Variscan Permian ‘red-bed’ sedimentary basins (Edwards et al., 1997) 165 and voluminous post-collisional felsic magmatism. The principal expression of 166 the latter is the Cornubian Batholith which comprises a variety of granite types 167 and is associated with rhyolite / microgranite dykes known locally as ‘elvans’ 168 (Simons et al., 2016). Host rocks within 1 km or so of the batholith margins 169 exhibit contact metamorphism. The granites, elvans and their surrounding 170 host rocks are cut by extensional fault systems hosting Early-Mid-Permian © British Crown Owned Copyright 2018/AWE 5

OFFICIAL

171 polymetallic W-Sn-Cu-As-Pb-Zn magmatic-hydrothermal mineralisation, the 172 working of which has resulted in substantial areas of metal-contaminated 173 made ground throughout the region (Pirrie et al., 2003). A subordinate Mid- 174 Triassic fault-controlled epithermal mineralisation episode, associated with the 175 migration of basinal brines from ‘red-bed’ basins, resulted in localized Pb-Zn- 176 barite mineralization (Simons et al., 2011). 177 The Quaternary geological history of SW England is dominated by 178 repeated intervals of periglacial climates and warmer inter-glacials. Although

179 some authors have speculated on the possibility of small cirque glaciers 180 (Harrison et al., 1998), there is no evidence for significant glaciation 181 throughout the region during the repeated cold climate intervals of the 182 Quaternary. Instead the bedrock geology was influenced by periglacial 183 processes with down-slope mass wasting of the local bedrock units. These 184 locally derived periglacial sediments are referred to as “head deposits” and 185 typically thicken into valleys. Along the north coast marine-derived carbonate 186 blown sand deposits are locally developed as dune sand systems. Flooded

187 steep sided valley systems known as rias, now forming major estuaries, are 188 developed along the south coast (e.g. the Helford, Fal, Fowey and Tamar 189 estuaries) along with smaller scale systems on the north coast (Hayle, Gannel 190 and Camel estuaries). These estuaries have received significant volumes of 191 contaminated mine waste as a result of the historical mining activity in the 192 region (Pirrie and Shail, 2018). 193 For the purposes of presentation and analysis of results, six broad 194 lithological groupings were identified from geological maps into which the

195 sampling locations were assigned. These were granitic (samples 1, 3, 10, 17, 196 20 and 34), mafic-igneous (samples 2, 12, 15, 22, 32, 35), metamorphic 197 (samples 4, 5, 14, 16), metamorphosed mudstone (samples 6, 7, 8, 13, 18, 198 19, 21, 23, 24, 26, 27, 28, 29), metamorphosed sandstone (samples 9, 11, 25) 199 and Carboniferous sedimentary (samples 30, 31 and 33). 200

201 Geography, climate and soils 202

OFFICIAL

205 agricultural land use. Lower lying areas are typically underlain by the 206 Devonian and Carboniferous metasedimentary rocks whilst the granites 207 underlay upland moorland which has an elevation of up to 621 m on 208 Dartmoor. The region has an oceanic climate, with a small range in annual 209 mean temperature between approximately 6ºC (winter) and 16ºC (summer), 210 mild winter temperatures and among the highest wind speeds in the United 211 Kingdom. Winters are wet and summers dry with annual precipitation totals of 212 900–1,000 mm at the coast, although annual precipitation increases inland, to

213 up to 2,000 mm on the granite moorland (Kosanic et al., 2014). 214 According to the available soil classification schemes, the soil types 215 present show relatively little variation within the studied area 216 (www.landis.org.uk/soilscapes). The majority of the Devonian and 217 Carboniferous metasedimentary rocks are overlain by acidic loamy soils, 218 slightly acidic loamy soils or slightly acidic base-rich soils. The granites are 219 overlain by the same acid loamy soils, along with very acid loamy upland soils 220 with a wet peaty surface, very acid upland soils with a peaty surface, and on

221 Bodmin Moor and Dartmoor blanket bog peat soils. Extensive areas on the St 222 Austell Granite are restored soils as a result of the china clay (kaolinite) 223 mining in this area. The soils on the Lizard Complex comprise slightly acidic 224 base-rich soils and wet acid loamy and clayey soils. 225 226 Materials and methods

227 228 Soil sampling

229 230 Prior to sampling, 35 rural locations throughout Devon and Cornwall 231 with (a) different underlying bedrock geology but (b) similar land-use, were 232 identified as suitable for sampling based on examination of topographic maps, 233 British Geological Survey geological maps and Google Earth imagery. Soil 234 sampling was carried out between the 24th February and the 7th March 2016. 235 Wherever possible, areas of managed grassland, ploughed and re-seeded for 236 grazing, were selected for sampling. Sampling sites within these areas were 237 selected away from sources of apparent introduced geological materials such 238 as tracks and roadways. Localised areas of historic mining activity are © British Crown Owned Copyright 2018/AWE 7

OFFICIAL

239 widespread in SW England; however, these were also deliberately avoided as 240 mineral contamination introduced into the surface environment near these 241 sites is highly distinctive, and therefore not representative of the underlying 242 bedrock-soil relationship (Pirrie et al., 2003). In the context of geolocation, 243 identification of these distinctive mineral species would allow an unknown 244 area to be identified with high precision; however, these signatures do not 245 have wider relevance in localities without a significant history of mining 246 contamination.

247 At each sampling location a 250 m long, linear transect was 248 established. In some cases, transects were offset to take into account the 249 orientation of field boundaries. Five samples were collected in total along the 250 line of the transect, with the first sample at the start (0 m), followed by 251 samples taken at distances of 10 m, 50 m, 100 m and 250 m along the 252 transect. Sampling location co-ordinates were recorded using a handheld 253 GPS. 254 To extract the soil sample, a labelled 35 mm diameter clean plastic

255 sampling pot was pushed approximately 2 cm into the ground surface (Fig. 2). 256 The edges of the pots are thin and effectively cut a clean edge into the soil 257 profile. If possible, areas of exposed soil were preferentially selected for 258 sampling. The in-situ sampling pots were digitally photographed, removed 259 and then sealed with tamper evident lids and placed within zip lock plastic 260 bags. The pots were then opened so that the exposed visible surface 261 represents the lowest part of the sampled soil, estimated to be at a depth of 262 between 1 and 2 cm below the surface. A subsample was removed from this

263 lower surface and dried for 2 hours at a temperature of 50°C, before being 264 sealed within a zip-lock plastic bag. In total, 175 soil samples were collected, 265 subsampled and dried. 266

267 Mineral analysis 268

269 The initial (0 m) sample from each of the 35 locations was selected for 270 mineral analysis (Table 2). In addition, one sampling location was selected at 271 random (location 24, Porthtowan Formation metamorphosed mudstone) and 272 all 5 samples collected from that location were also prepared for analysis. © British Crown Owned Copyright 2018/AWE 8

OFFICIAL

273 Each soil subsample was gently disaggregated, placed into a 30 mm mould 274 and mixed with epofix resin. The samples were allowed to cure for 24 hrs, 275 labelled and back-filled with araldite resin and heated at 50°C for 2 hours. 276 The blocks were then polished and carbon coated prior to mineral analysis. 277 Mineral analysis of the 40 samples was carried out using automated scanning 278 electron microscopy (SEM) with linked energy dispersive X-ray (EDX) 279 spectrometers, based on QEMSCAN technology. 280 Automated SEM-EDS is a widely used method for mineral analysis

281 (Armitage et al., 2010; Williamson et al., 2013; Eby et al., 2015) and has 282 previously been used for the forensic analysis of soil samples (Pirrie and 283 Rollinson, 2011; Pirrie et al., 2004, 2013, 2014). In brief, individual particles 284 are located on the cut face of the polished block and are then phase mapped 285 in cross section by the acquisition of energy dispersive X-ray spectra (in this 286 study using 1000 counts/pixel) at a regular, operator defined spacing across 287 the sample. The 1000 count spectra places limits of detection for elements 288 present within the individual mineral grain at an abundance of approximately 3

289 atom percent (Andersen et al., 2009). In this study spectra were acquired 290 across the sample using a 6 µm spacing, and the particle size range accepted 291 for analysis was 9 µm to 800 µm. Each measured spectrum is assigned to a 292 mineral name or chemical grouping by matching the elemental signature 293 against a user-defined classification scheme (Pirrie and Rollinson, 2011). The 294 mineral categories used in this study are provided in Table 3. 295 In each soil sample, >5000 individual mineral grains were analysed, 296 based on the acquisition of between 50,734 and 487,685 EDS analysis points

297 per sample. The number of EDS analyses is controlled by the size of the 298 individual soil particles, such that finer-grained particles are effectively 299 mapped by a smaller number of spectra than a larger particle (Table 2). Each 300 sample took between 20 and 60 minutes to be analysed, showing that the 301 methodology is sufficient for rapid sample screening. The automated 302 mineralogy data were acquired using an FEI 650F scanning electron 303 microscope, with an accelerating voltage of 25kv, equipped with twin Bruker 304 30 mm2 Quantax spectrometers, and the data were processed using iDiscover 305 5.3 software. The data outputs derived from the automated mineral analysis

OFFICIAL

306 used in this study are: modal mineralogy, mean mineral particle size and 307 QEMSCAN mineral and phase composition particle maps. 308 309 Results 310 311 In the following description the mineral abundance is referred to as 312 major >10%, minor 1-10% and trace <1% abundance in the measured 313 sample.

314 315 Soil mineralogy variation

316 317 The variation in the modal mineralogy for the 35 different sampling 318 locations across SW England is shown in Figs. 3a – f, and tabulated in Table 319 4. Although there is a range in bedrock geology in the study area the majority 320 of the soil samples analysed are dominated by the same principal mineral 321 types: quartz, plagioclase, K-feldspar, muscovite, biotite, chlorite, kaolinite, Fe

322 silicates, Mg silicates, hornblende and tourmaline (Fig. 3a, c and e, Table 4). 323 Less abundant minerals present are: epidote group minerals, sillimanite, 324 topaz, other silicates, calcite, rutile, ilmenite, Fe-Mn oxides, chromite, apatite, 325 zircon, xenotime, Ce phosphates, Fe/Cu sulphides, gypsum and “others” (Fig. 326 3b, d, and f, Table 4). However, although the same major/minor and trace 327 minerals tend to co-occur throughout the samples analysed, their relative 328 abundance varies considerably (Fig. 3a). When the data are combined by the 329 location rock type as described in Table 1, differences both between and

330 within these groups are observed. Although there is some mineralogical 331 variation within each lithological group, there is greater variation in modal 332 mineralogy between the different lithological groupings than observed within 333 an individual group. These differences between lithological groups indicate the 334 potential for the ‘type’ of geology to be identified in an unknown sample. The 335 observed difference detected between samples of the same lithology 336 highlights that, within the study area, there is potential for individual locations 337 to be identified by their particular signatures. 338 In addition to the modal mineralogical dataset, QEMSCAN particle 339 compositional images provide further means of classifying differences © British Crown Owned Copyright 2018/AWE 10

OFFICIAL

340 between the analysed soils. These reveal textural relationships between 341 different mineral phases in the samples relating to grain size and mineral 342 association, which can vary depending on the bedrock lithology on which the 343 soil was developed. Representative particle images for selected soil samples 344 developed on the different bedrock groups are shown in Fig. 4. Clear 345 differences are observed in the mineral textures for soils developed on 346 different rock types. For example, soils underlain by granitic bedrock include 347 coarse grains of quartz, feldspar and mica minerals, whereas soil particles

348 derived from metasedimentary rocks are highly heterogeneous, principally 349 comprising grains of quartz intermixed with fine-grained clay minerals. No two 350 soil samples analysed contain the same mineral types with the same relative 351 abundance, grain size and texture, reflecting the distinctive nature of soil 352 composition at different locations. The differences in characteristics between 353 soils developed on different underlying bedrock, which are evident in 354 groupings of automated mineralogy data (Fig. 3), could allow broad areas to 355 be identified or ruled out by identifying the likely source lithologies from an

356 unknown sample. 357

358 Soils grouped by bedrock lithology 359

360 To assess the mineralogical variation throughout the study area in 361 more detail the data for the same, or similar, bedrock geology is compared. 362 Identification of differences in soil mineralogy between samples arising from 363 similar bedrock geologies is required to define the range of compositions

364 which could be indicative of a particular lithology. If specific differences are 365 present, these could also be used to further refine the spatial scale of a 366 geolocation assessment, such as identifying a particular group out of several 367 candidate units within a region. 368 369 Granites. Six soil samples developed upon granite bedrock were analysed. 370 These soils were identified as being derived from an underlying granitic 371 geology by the consistent abundance of large (~100 µm) grains of quartz, 372 plagioclase, K feldspar, biotite and muscovite (e.g. Fig. 4). The Cornubian 373 granites have been subdivided into five mineralogical types and individual © British Crown Owned Copyright 2018/AWE 11

OFFICIAL

374 named intrusions are commonly composite with more than one granite type 375 present (Simons et al., 2016). Each of the soils sampled had distinctive 376 compositions in minor and trace mineralogy, which were consistent with at 377 least one of the characteristics of the underlying subcrop. This is indicative of 378 the known spatial heterogeneity of the individual granite bodies (Simons et al., 379 2016). 380 Key mineralogical parameters and the known underlying granite types 381 are presented in Table 5. The relative abundance of biotite and muscovite

382 was indicative of biotite granite (G3, biotite > muscovite) underlying the Lands 383 End granite soil, and two-mica granite (G1, biotite ≈ muscovite) underlying the 384 Tregonning-Godolphin and Carnmenellis Granite soils. High relative 385 abundance of tourmaline in the St Austell and Dartmoor samples was 386 indicative of G4 tourmaline granites in these sample areas. The Bodmin Moor 387 Granite was characterised by an increased abundance of muscovite relative 388 to biotite, consistent with muscovite granite (G2), and the abundance of 389 ‘topaz’ and ‘sillimanite’ (suggestive of the accessory mineral andalusite

390 (Simons et al., 2016), which would be grouped with sillimanite during 391 QEMSCAN analysis) was distinctive in this sample. Other mineralogical 392 signatures in the group included kaolinite abundance, which was particularly 393 high in the St Austell Granite soil. Replacement of feldspar by kaolinite is 394 present in all of the granites of SW England, however, the St Austell Granite 395 exhibits the most extensive kaolinisation (Ellis and Scott, 2004), and the high 396 abundance in this sample is evidence that this mineralogical signature is 397 preserved in the soil.

398 Within the study area, soils developed on granites possess both 399 common overall characteristics that allows the underlying lithology to be 400 identified, but also sufficient variety between samples to permit individual 401 subcrops to be distinguished. Links between the composition of a soil and 402 published characterisation of the granite type (Ellis and Scott, 2004; Simons et 403 al., 2016) were possible, such that key mineral signatures can be used to rule 404 out one or more of the subcrops as an unlikely source of the soil. This 405 approach is powerful, but may be highly limited to locations or rock types 406 which display this degree of mineralogical variation. 407 © British Crown Owned Copyright 2018/AWE 12

OFFICIAL

408 Mafic igneous rock. Six soil samples were analysed from locations where the 409 reported bedrock geology was gabbro, dolerite or basalt (Table 3), and were 410 grouped together as ‘mafic igneous’ rock types. Modal mineralogy data (Fig 411 3a and 2b) shows that although there is some variation between these 412 samples, they form a consistent grouping characterised by high abundance of 413 chlorite and the presence of Mg silicates, both of which are indicative of 414 altered mafic minerals. 415 Soils developed on dolerites and gabbros had similar overall

416 composition and texture (e.g. Fig 4), comprising several species of coarse 417 mono-minerallic grains suggestive of an igneous basement rock. The samples 418 are however, subtly different to each other, with the gabbro derived soils 419 containing more abundant K feldspar, hornblende and kaolinite, and less 420 abundant Mg silicates, chlorite and Fe silicates than the dolerites. Highly 421 specific mineral signatures were detected in some samples, in particular the 422 presence of abundant hornblende in Sample 15, which within the study area 423 was distinctive of the Trelan and Crousa Gabbro, part of the Lizard Complex.

424 The two soil samples from locations underlain by either basaltic 425 calcareous tuffs of the Tintagel Volcanic Formation or the basalts-mudstones 426 of the Milton Abbott Formation were distinct from the soils derived from 427 dolerite/gabbro bedrock. In general, the soils associated with basaltic 428 lithologies contain less abundant plagioclase, hornblende and tourmaline and 429 more abundant K feldspar and biotite, although it should be noted that the 430 basaltic lithologies are interbedded with mudstones. In addition, whilst the 431 Tintagel Volcanic Formation is described as comprising calcareous basaltic

432 tuffs, very few (0.10%) grains in the soil sample derived from this unit reported 433 to the calcite mineral grouping. Geochemical and mineralogical studies have 434 indicated that the Tintagel Volcanic Formation comprises both calcite-rich but 435 also calcite-poor lithologies (Rice-Birchall and Floyd, 1988), hence it is 436 possible that the sample collected was from soils developed on a calcite-poor 437 lithological unit. 438 The soils underlain by mafic-igneous lithologies were found to form a 439 consistent grouping of modal mineralogy and texture. Owing to the range of 440 rock types grouped together, aspects of the mineralogy of both individual 441 samples and rock types (e.g. gabbro vs dolerite) were found to be distinctive. © British Crown Owned Copyright 2018/AWE 13

OFFICIAL

442 In contrast to granitic derived soils, the mineralogy of these samples is more 443 strongly controlled by alteration products of mafic minerals, rather than the 444 persistence of lithic fragments. 445 446 Metamorphic rocks of the Lizard Complex. Four soil samples collected from 447 locations underlain by metamorphic rocks were analysed, all associated with 448 the Lizard Complex (Fig. 1). It should be noted that this grouping of soil 449 samples is rather arbitrary, as it does not necessarily represent a

450 homogenous compositional grouping but could include rocks of differing 451 mineralogy depending on protolith composition and metamorphic grade. 452 Some of the samples showed similar major mineralogy to locations within the 453 mafic-igneous group, however, they were identified as metamorphic derived 454 soils by the enhanced abundance of mineral groups such as hornblende, 455 epidote and Mg-silicates and a lower abundance of rutile and ilmenite. 456 Trends relating to specific units within the Lizard Complex were 457 identified. Soil sample 14 has the highest abundance of hornblende (12.8%)

458 and epidote (1.6%) of the entire sample set. Within the sample area this 459 combination of hornblende and epidote, a common greenschist-facies 460 metamorphic mineral, is distinctive of the Traboe Hornblende Schist 461 underlying this sample location. 462 The abundance of Mg-silicate minerals was expected to be a 463 distinguishing characteristic of soils underlain by the Lizard Serpentinite, as 464 this is the compositional grouping which includes serpentine-group minerals 465 (Table 3). However, although the soil from this area (Sample 16) contained

466 1.4% Mg-silicates (serpentine group minerals), two samples from other areas 467 contained higher abundances; samples 4 (11.1% Mg-silicates, gneiss) and 12 468 (3.9% Mg-silicates, dolerite). The published geological map indicated that 469 sampling location 4 was underlain by the Kennack Gneiss. However, this 470 lithology has a patchy / localised distribution with pods of granitic gneiss 471 surrounded by serpentinite. Consequently, the soil mineralogy data may be 472 interpreted to suggest that either: (a) the published map is incorrect, and that 473 the bedrock geology at this location is actually serpentinite, or that (b) 474 periglacial processes have introduced serpentinite derived surficial deposits at 475 this location. It should however, also be noted that both the Kennack Gneiss © British Crown Owned Copyright 2018/AWE 14

OFFICIAL

476 (Sample 4) and Nare Head Dolerite (Sample 12) contain amphibole group 477 minerals (Barnes et al., 1979), which (other than hornblende) would be 478 characterised under the Mg silicate QEMSCAN grouping. Within the study 479 area, soils with abundant Mg-silicate (>1%) were consistent with several 480 underlying lithologies, however, this was highly distinctive, relating to three out 481 of the thirty five locations sampled. More generally, this finding demonstrates 482 the importance of obtaining detailed information on the composition of 483 candidate lithologies to check for consistency with observed soil mineralogy.

484 Reference to regional soil geochemistry surveys revealed that a 485 characteristic feature of the soils present on the Lizard Complex is a high 486 relative abundance of chromium (Cr) (BGS, 2015). This was supported in the 487 modal mineralogy data with elevated abundances of the mineral chrome 488 spinel (Table 2) for soils developed on the Lizard Serpentinite (Sample 16) 489 and also at sample location 4. The abundance of chrome spinel along with 490 the serpentine minerals would support the interpretation that sampling location 491 4 is not underlain by the Kennack Gneiss as indicated by the published map,

492 but instead is underlain by serpentinite. This therefore provides a highly 493 location specific signature which is absent in soils developed on chromium- 494 poor rocks elsewhere in the sample region. 495 496 Metamorphosed mudstone. The dominant geological unit throughout the 497 study area is Devonian low-grade metamorphosed mudstones and 498 sandstones, the original deposition of which was strongly controlled by 499 tectonics, within a series of separate depositional basins in the study area;

500 Gramscatho, Looe, South Devon and Tavy (Fig. 1) (Leveridge, 2011; 501 Leveridge and Shail, 2011b). 502 Thirteen soil samples developed upon Devonian metamorphosed 503 mudstones were analysed, covering each of the depositional basins (Fig. 3c, 504 d). These samples had a characteristic fine-grained texture (Fig. 4) and 505 similar modal mineralogy, being dominated by major quartz, K feldspar, 506 muscovite, biotite and chlorite along with minor/trace plagioclase, kaolinite 507 and tourmaline. This combination of features permitted distinction of these 508 samples as being derived from a metasedimentary source.

OFFICIAL

509 Despite the similar overall characteristics, it is clear that there are 510 mineralogical variations within the dataset related to depositional basin (Fig. 511 3c, d). For example, soil samples collected from locations up to ~60 km apart 512 across the South Devon Basin (19, 23, 28 and 29) are all quartz-poor (less 513 than 20% quartz, Fig. 3c), whilst those collected from the Gramscatho Basin 514 (13, 18, 24, 26), again up to 60 km apart, have a consistently higher quartz 515 abundance (>20-30%). The changing relative abundance of quartz may 516 reflect differences in soil particle size, as quartz typically comprises larger

517 sand/silt sized particles. However, the average particle size for Gramscatho 518 Basin samples was finer than in the South Devon Basin samples, and 519 therefore the quartz variations are likely to be representative of mineralogical 520 differences between the locations rather than particle size. 521 In the soil mineralogy data, samples collected from the Trevose Slate 522 (Sample 19, Devon Basin) and the Tredorn Slate (Sample 21, Tavy Basin) 523 formations both have elevated levels of chrome spinel (Table 2). Although 524 this is not described in regional scale geological mapping, published soil

525 geochemical surveys show an elevated chromium anomaly in these regions 526 (BGS, 2015) consistent with the measured mineralogical data. Whilst Cr-rich 527 soils are observed in other regions of the study area such as the Lizard 528 Complex, a soil with meta-mudstone characteristics and the presence of Cr- 529 spinel as a minor mineral indicates a restricted provenance to the Trevose- 530 Tredorn formations. 531 532 Metamorphosed sandstone-mudstone. Three soil samples were analysed

533 from locations underlain by either metamorphosed Devonian sandstones or 534 interbedded sandstones / mudstones (Figs. 3e and f). These were 535 differentiated from the metamorphosed Devonian mudstones principally by 536 texture; soils developed on lithologies with a sandstone component contained 537 quartz as both sand-silt grade monominerallic and sand-grade polyminerallic 538 grains (Fig. 5). The abundance of quartz is another key differentiator. Sample 539 9, which was taken from a sandstone bedrock location has very high quartz 540 abundance (73.7%), whilst soils from interbedded locations (samples 11 and 541 25) are intermediate in quartz abundance between sandstone and mudstone 542 locations. In these cases, the higher abundance of quartz is indicative of the © British Crown Owned Copyright 2018/AWE 16

OFFICIAL

543 coarser grain size of the underlying rock at these locations, which along with 544 grain texture, may allow the identification of soils developed on different grain 545 size sedimentary or metasedimentary rocks. 546 547 Carboniferous sedimentary rocks. Sedimentary rocks of Carboniferous age 548 which have undergone Variscan deformation and very low grade regional 549 metamorphism occur within the eastern part of the study area. Three soil 550 samples developed on Carboniferous mudstones, interbedded mudstones

551 and sandstones, and chert were analysed (Fig. 3e and f). Soils developed on 552 locations underlain by Carboniferous mudstone and chert formations 553 (Samples 30 and 31) had similar fine-grained textures to the Devonian 554 metamorphosed mudstone samples (e.g. Fig. 4 d and f). However, the lower 555 metamorphic grade of the Carboniferous rock resulted in mineralogical 556 differences, with lower biotite and muscovite (illite) abundances and more 557 abundant chlorite, plagioclase and quartz reflecting lower thermal maturity of 558 these bedrocks. The abundance of quartz in the ‘chert’ sample was not as

559 high as may be expected, and therefore may arise from another interbedded 560 component (e.g. mudstone) of the heterogeneous Teign (Newton) Chert 561 Formation. 562 The sandstone-mudstone nature of Sample 33 could be readily 563 identified amongst the Carboniferous sedimentary samples by the high 564 abundance of quartz and coarse grain size. However, it could not be readily 565 distinguished from the Devonian metamorphosed location (Sample 9); unlike 566 the mudstones, little mineralogical change would be expected during the low

567 grade metamorphism of a quartz-rich sandstone and therefore no difference 568 between these samples was detected. This finding highlights that using 569 automated mineralogy alone it may be difficult to distinguish between lower 570 grade metamorphosed sediments and ‘pure’ sedimentary lithologies, and in 571 areas predominantly composed of such rocks a wider suite of techniques will 572 be required.

573 574 Variation within a single location

575

OFFICIAL

576 To examine the spatial variation at a more local scale in detail, one 577 sample was selected at random and all five sampling points along a 0 to 250 578 m transect were prepared for analysis. The modal mineralogy for these five 579 samples, along with the replicate sample analysed from the first sampling 580 point are illustrated in Fig. 6 and the data are presented in Table 6. The 581 mineralogical data show the closest correspondence between the two 582 replicate samples, but the overall suite of five samples collected over a 583 distance of 250 m within the same agricultural field are a closely comparable

584 data set. The data for the five samples along the individual transect show 585 lower variance than the samples analysed from the different sampling 586 locations. Thus if the data for the 35 separate sampling locations across the 587 region are compared with the data for these samples, then it is clear that the 588 soil modal mineralogical data would allow the identification of the correct 589 sampling location. This supports the modal mineralogy data from the whole 590 sample suite, which indicates that within the study area, the underlying 591 bedrock geology is the dominant control on soil mineral composition.

592 593 Discussion 594 595 For forensic geolocation based on a soil sample, the ideal scenario 596 would be to have access to representative samples from all the regions of 597 interest to allow comparisons to be drawn directly. However, this is often not 598 practical, and may be impossible in wide-area search investigations, which 599 would be characteristic in nuclear forensics. Therefore, it is necessary to rely

600 on comparison of measurable properties of an unknown soil with reference 601 datasets to establish the potential provenance of a sample. 602 Most spatial soil surveys typically describe parameters such as grain 603 size, organic content, pH, colour etc, which for geolocation purposes are not 604 sufficiently descriptive. For example, soil surveys from the study area show 605 little variation other than the major contrasts between upland peat soils versus 606 the predominantly managed arable and grazing land. This contrasts markedly 607 with the clear differences in soil mineralogy and particle textures revealed by 608 automated mineralogical analysis, and this variety indicates that, within the

OFFICIAL

609 studied area, soil mineralogy is a more distinctive indicator of location than the 610 overall soil classification or texture. 611 There are few comprehensive global datasets for soil mineralogy. The 612 United States Geological Survey has an open access database for soil 613 mineralogy of the USA, comprising 4857 samples, which equates to a 614 sampling density of 1 sample per 1600 km2 (Smith et al., 2014). As such this 615 shows regional trends, but is of less value for forensic investigations. In the 616 absence of comprehensive datasets for detailed soil mineralogy, other

617 reference datasets must be used to predict locations at which a particular soil 618 mineralogy may be found. In this paper we have demonstrated that, for the 619 study area, soil mineralogy is consistent with the local underlying geology as 620 represented by widely available regional scale geological mapping. Local 621 scale geochemical surveys (BGS, 2015) can add an extra level of detail, such 622 as the highly localised distribution of chromium around the Trevose and 623 Tredorn Slate formation locations, which was detected as a high relative 624 abundance of chrome-spinel in these soils samples.

625 The nature of the correlation between bedrock composition and soil 626 mineralogy will depend on a number of factors specific to the study area. In 627 localities dominated by physical weathering, with typically arid to semi-arid 628 climates, soils are more likely to be composed of liberated relic minerals grain 629 and lithological fragments, and it is expected that there will be a strong 630 correlation between the mineralogical composition of soils and the near- 631 surface geology. However, in areas where chemical weathering 632 predominates, and/or where soil profiles relate to past intervals of geological

633 weathering, soil mineralogy may be substantially altered from the parent rock. 634 Owing to the oceanic climate, soil formation in the study area is dominated by 635 chemical weathering and hence is more representative of the second 636 condition. 637 Soil mineralogy was also sufficient to distinguish individual occurrences 638 of most of the lithology types within the study area. Consideration of the minor 639 phases present in the granites showed differences related to the specific gran- 640 ite type whereas the mineralogy of metasediment samples was shown to differ 641 across different depositional basins. The extent to which this might be used to 642 determine a specific location, rather than just distinguish two dissimilar sam- © British Crown Owned Copyright 2018/AWE 19

OFFICIAL

643 ples, was varied and depended on the mineralogical variety within the parent 644 rock. It was only in the case of the sandstone samples 9 and 33 that these 645 distinctions could not be made; including between different metamorphosed 646 grades of this lithology. Nevertheless, these results highlight the potential of a 647 two-stage approach to predictive geolocation, whereby a generic lithology 648 may be identified in the first instance, and, in favourable locations, further re- 649 finement made by considering the details of local variations in specific occur- 650 rences of that lithology. Whilst regional scale mapping is probably sufficient for

651 the first aspect, more detailed data sources, such as specific publications, 652 may be required for the second. 653 In the context of nuclear forensics, the capability provided by soil 654 analysis to identify even the generic geological background may be of 655 particular value. In many cases, such data may permit some potential 656 locations to be excluded, rather than indicating an individual specific target 657 location. The significance of this is that during a nuclear forensic 658 investigation, the nature of a material being examined may indicate that it

659 could only have been derived from a limited number of potential installations. 660 Consideration of geolocation evidence (where present) may therefore allow 661 some of these candidate sources to be ruled out based on their environmental 662 setting, substantially narrowing the scope of the investigation. Further work to 663 consider the collection and analysis of this type of evidence in nuclear forensic 664 scenarios is needed. Although it may not be expected to recover soil traces in 665 all cases, the power of this analysis to distinguish locations, if soil is present, 666 highlights its value to these complex investigations.

667 668 Conclusions 669 670 In this study we aimed to test the extent to which soil mineralogy can 671 be used to predict an underlying geological rock type, and how distinctive this 672 may be of an individual unit of that lithology. The study area in SW England 673 may represent a ‘worst-case’ type environment for this technique, as soils in 674 the area are highly chemically altered from the host rock mineralogy through 675 chemical weathering and human activity. The repetition of many of the major 676 rock types throughout the area also challenged the capability of this analysis © British Crown Owned Copyright 2018/AWE 20

OFFICIAL

677 to distinguish different locations with similar underlying rock types. In all 678 sampled locations, signatures of the underlying bedrock geology were 679 detected in the soil mineralogy and particle textures, and in 33 out of the 35 680 samples, location-specific were detected permitting different units of the same 681 major rock type to be distinguished. 682 This type of analysis is a powerful method for the identification of 683 potential underlying bedrock units from soil samples, which can be used with 684 readily available regional scale geological mapping to identify, and just as

685 importantly rule out, candidate regions based on the known geology. The 686 importance of this information will depend on the extent of geological variation 687 within the area of investigation; if the entire area is underlain by the same 688 lithology then other geolocation indicators may be more valuable. 689 In our area of interest in nuclear forensics, the provenance of a found 690 nuclear material may be already somewhat constrained by known holdings 691 and operations, and therefore identification of the geology type may allow 692 some potential sources to be ruled out. Challenges relating to transfer and

693 collection of soil evidence in these scenarios must be evaluated, however, 694 consideration of environmental signatures such as soil evidence has the 695 potential to provide a unique insight and perspective in these complex 696 investigations. 697 698 Acknowledgements 699 700 We thank Holly Campbell (Helford Geoscience LLP) for assistance with the

701 soil sampling and preparation of samples for analysis. Figure 1B reproduced 702 from R.K. Shail and B.E. Leveridge. The Rhenohercynian passive margin of 703 SW England: Development, inversion and extensional reactivation. Comptes 704 Rendus Geoscience, 2009, 341 (issue2-3), 140-155. Copyright (c) 2009 705 published by Elsevier Masson SAS. All rights reserved. Funding for this 706 programme of research was provided by AWE plc to Helford Geoscience LLP. 707 708 References 709

OFFICIAL

710 Anderson, J.C.O., Rollinson, G.K., Snook, B., Herrington, R. & Fairhurst, R.J. 711 2009. Use of QEMSCAN for the characterisation of Ni-rich and Ni-poor 712 goethite in laterite ores. Minerals Engineering, 22, 1119-1129. 713 Armitage, P.J., Worden, R.H., Faulkner, D.R., Aplin, A.C., Butcher, A.R. & 714 Iliffe, J. 2010. Diagenetic and sedimentary controls on porosity in 715 Lower Carboniferous fine-grained lithologies, Krechba field, Algeria: A 716 petrological study of a caprock to a carbon capture site. Marine and 717 Petroleum Geology, 27, 1395-1410.

718 Barnes, R.P., Andrews, J.R. & Badham, J.P.N. 1979. Preliminary 719 investigations of South Cornish melanges. Proceedings of the Ussher 720 Society, 4(3), 262-268. 721 Beamish, D. 2015. Relationships between gamma-ray attenuation and soils in 722 SW England. Geoderma, 259-260, 174-186. 723 Benham, A.J., McEvoy, F.M. & Rollin, K.E. 2005. Potential for stratiform 724 massive sulphide mineralisation in south-west England. Transactions 725 of the Institute of Mining and Metallurgy, Section B, 113, 227-246.

726 British Geological Survey, NERC. 2015. Geochemical baseline survey of the 727 environment, South West England. 728 Bowen, A.M. & Caven, E.A. 2013. Forensic provenance investigations of soil 729 and sediment samples. In: Pirrie, D., Ruffell, A. & Dawson, L.A. (eds) 730 Environmental and Criminal Geoforensics. Geological Society, 731 London, Special Publication, 384, 9-25. 732 Bull, P.A., Parker, A. & Morgan, R.M. 2006. The forensic analysis of soils and 733 sediment taken from the cast of a footprint. Forensic Science

734 International, 162, 6–12. 735 Caritat, P., Simpson, T. & Woods, B. 2019. Predictive soil provenancing 736 (PSP): an innovative forensic soil provenance analysis tool. Journal of 737 Forensic Sciences. doi: 10.1111/1556-4029.14060. 738 Cranfield Soil and Agrifood Institute, soilscapes database, last accessed 739 21/09/2016, http://www.landis.org.uk/soilscapes. 740 Eby, G.N., Chamley, N., Pirrie, D., Hermes, R., Smoglia, J. & Rollinson, G. 741 2015. Trinitite redux: Mineralogy and petrology. American Mineralogist, 742 100, 427-441.

OFFICIAL

743 Edwards, R.A., Warrington, G., Scrivener, R.C., Jones, N.S., Haslam, H.W., 744 Ault, L. 1997. The Exeter Group, south Devon, England: a contribution 745 to the early post-Variscan stratigraphy of northwest Europe. Geological 746 Magazine, 134, 177-197. 747 Ellis, R.J. & Scott, P.W. 2004. Evaluation of hyperspectral remote sensing as 748 a means of environmental monitoring in the St. Austell china clay 749 (kaolin) region, Cornwall, UK. Remote Sensing of the Environment, 93, 750 118-130.

751 Fitzpatrick, R., Raven, M.D. & Forrester, S.T. 2009. A systematic approach to 752 soil forensics: Criminal case studies involving transference from crime 753 scene to forensic evidence. In: Ritz, K., Dawson, L. & Miller, D. (eds), 754 Criminal and Environmental Soil Forensics. Springer, 105-127. 755 Harrison, S., Anderson, E. & Passmore, D.G. 1998. A small glacial cirque 756 basin on Exmoor, Somerset. Proceedings of the Geologists 757 Association, 109, 149-158. 758 Keegan, E., Kristo, M.J., Toole, K., Kips, R. & Young, E. 2016. Nuclear

759 Forensics: Scientific Analysis Supporting Law Enforcement and 760 Nuclear Security Investigations. Analytical Chemistry, 88, 1496-1505. 761 Kirkwood, C., Everett, P., Ferreira, A. & Lister, B. 2016. Stream sediment 762 geochemistry as a tool for enhancing geological understanding: An 763 overview of new data from south west England. Journal of 764 Geochemical Exploration, 163, 28-40. 765 Kosanic, A., Harrison, S. Anderson, K. & Kavcic, I. 2014. Present and 766 historical climate variability in South West England. Climate Change,

767 124, 221-237. 768 Lark, R.M. & Rawlins, B.G. 2008. Can we predict the provenance of a soil 769 sample for forensic purposes by reference to a spatial database? 770 European Journal of Soil Science, 59, 1000-1006. 771 Leveridge, B.E. 2011. The Looe, South Devon and Tavy basins: the 772 Devonian rifted passive margin successions. Proceedings of the 773 Geologists Association, 122, 616–717. 774 Leveridge, B.E. & Shail, R.K. 2011a. The Gramscatho Basin, south Cornwall, 775 UK: Devonian active margin successions. Proceedings of the 776 Geologists Association, 122, 568–615. © British Crown Owned Copyright 2018/AWE 23

OFFICIAL

777 Leveridge, B.E. & Shail, R.K. 2011b. The marine Devonian stratigraphy of 778 GB. Proceedings of the Geologists Association, 122, 540-567. 779 Mayer, K., Wallenius, M. & Varga, Z. 2012. Nuclear forensic science: 780 Correlating measureable material parameters to the history of nuclear 781 material. Chemical Reviews, 13, 884 - 900. 782 Mayer, K., Wallenius, M., Lützenkirchen, K., Horta, J., Nicholl, A., Rasmussen, 783 G., van Belle, P., Varga, Z., Buda, R., Erdmann, N., Kratz, J.-V., 784 Trautmann, N., Fifield, K., Tims, S., Fröhlich, M. & Steier, P. 2015.

785 Uranium from German Nuclear Power Projects of the 1940s – A 786 Nuclear Forensic Investigation. Angewandte Chemie, International 787 Edition, 54, 13137. 788 Owens, P.N., Blake, W.H., Gaspar, L., Gateuille, D., Koiter, A.J., Lobb, D.A., 789 Petticrew, E.L., Reiffarth, D.G., Smith, H.G. & Woodward, J.C. 2016. 790 Fingerprinting and tracing the source of soils and sediments: Earth and 791 ocean science, geoarchaeological, forensic and human health 792 implications. Earth Science Reviews, 162, 1-23.

793 Pirrie, D & Rollinson, G.K. 2011. Unlocking the application of automated 794 mineralogy. Geology Today, 27, 226-235. 795 Pirrie, D. & Ruffell, A. 2012. Soils and geological trace evidence. In: Marquez 796 Grant, N. and Roberts, J. (eds) Forensic Ecology Handbook. 797 Blackwell-Wiley, 183-201. 798 Pirrie, D. & Shail, R.K. 2018. Mud and metal; the impact of historical mining 799 on the estuaries of SW England, UK. Geology Today, 34, 215-223. 800 Pirrie, D., Power, M.R., Rollinson, G., Camm, G.S., Hughes, S.H., Butcher,

801 A.R. & Hughes, P. 2003. The spatial distribution and source of arsenic, 802 copper, tin and zinc within the surface sediments of the Fal Estuary, 803 Cornwall, UK. Sedimentology, 50, 579-595. 804 Pirrie, D., Butcher, A.R., Power, M.R., Gottlieb, P. & Miller, G.L. 2004. Rapid 805 quantitative mineral and phase analysis using automated scanning 806 electron microscopy (QemSCAN); potential applications in forensic 807 geoscience. In: Pye, K. & Croft, D. (eds), Forensic Geoscience. 808 Geological Society, London, Special Publication, 232, 123-136. 809 Pirrie, D., Rollinson, G.K., Power, M.R. & Webb, J. 2013. Automated forensic 810 soil mineral analysis; testing the potential of lithotyping, In: Pirrie, D., © British Crown Owned Copyright 2018/AWE 24

OFFICIAL

811 Ruffell, A. & Dawson, L.A. (eds) Environmental and criminal 812 geoforensics. Geological Society of London, Special Publication, 384, 813 47-64. 814 Pirrie, D., Rollinson, G.K., Andersen, J.A., Wootton, D. & Moorhead, S. 2014. 815 Soil forensics as a tool to test reported artefact find sites. Journal of 816 Archaeological Science, 41, 461-473. 817 Pirrie, D., Dawson, L. & Graham, G. 2017. Predictive geolocation: forensic 818 soil analysis for provenance determination. Episodes, 40, 141-147.

819 Rice-Birchall, B. & Floyd, P.A. 1988. Geochemical and source characteristics 820 of the Tintagel Volcanic Formation. Proceedings of the Ussher Society, 821 7, 52-55. 822 Shail, R.K. & Leveridge, B.E. 2009. The Rhenohercynian passive margin of 823 SW England: development, inversion and extensional reactivation. 824 Comptes Rendus Geoscience, 341, 140-155. 825 Simons, B., Pirrie, D., Rollinson, G.K. & Shail, R.K. 2011. Geochemical and 826 mineralogical record of the impact of mining on the Teign estuary,

827 Devon, UK. Geoscience in South-West England, 12, 339-350. 828 Simons, B., Shail, R.K. & Andersen, J.C.O. 2016. The petrogenesis of the 829 Early Permian Variscan granites of the Cornubian Batholith: Lower 830 plate post-collisional peraluminous magmatism in the Rhenohercynian 831 Zone of SW England. Lithos, 260, 76-94. 832 Smith, D.B., Cannon, W.F., Woodruff, L.G., Solano, Federico & Ellefsen, K.J., 833 2014. Geochemical and mineralogical maps for soils of the 834 conterminous United States. U.S. Geological Survey Open-File Report,

835 2014–1082, 386 p. https://dx.doi.org/10.3133/ofr20141082. 836 Stern, L.A., Webb, J.B., Willard, D.A., Bernhardt, C.E., Korejwo, D.A., Bottrell, 837 M.C., McMahon, G.B., McMillan, N.J., Schuetter, J.M. & Hietpas, J. 838 2019. Geographic attribution of soils using probabilistic modeling of 839 GIS data for forensic search efforts. Geochemistry, Geophysics, 840 Geosystems, 20, 1-20. 841 Wallenius M., Mayer, K. & Ray, I. 2006. Nuclear forensic investigations: Two 842 case studies. Forensic Science International, 156, 55-62.

OFFICIAL

843 Williamson, B.J., Rollinson, G. & Pirrie, D. 2013. Automated mineralogical 844 analysis of PM10: New parameters for assessing PM toxicity. 845 Environmental Science and Technology, 47, 5570-5577. 846

847

OFFICIAL

848 Figure captions 849 850 Fig. 1. Simplified maps showing distribution of sampling sites (A) and major 851 near-surface geological features of SW England (B). Map B shows the main 852 Devonian sedimentary basins (Gramscatho, Looe, South Devon and Tavy 853 basins) along with the Lizard Ophiolite Complex (purple), the Start Complex 854 and the exposed granites making up the Cornubian Batholith. Map A based 855 on a Digimap extract © Crown Copyright and Database Right [June 2018].

856 Ordnance Survey (Digimap Licence). Map B from Shail and Leveridge 857 (2009). 858 859 Fig. 2. Example images for soil sampling location 1, near Halwyn Farm, 860 Mousehole, West Cornwall, underlain by the Lands End Granite. Five soil 861 samples were collected from the managed grassland field (A) at sampling 862 positions of 0 (1/1), 10 (1/2), 50 (1/3), 100 (1/4) and 250 m (1/5) along the 863 transect (B). Soil samples were collected by inserting a clean, single use

864 plastic vial into the soil to a depth of 1-2 cm (C). Map A based on a Digimap 865 extract © Crown Copyright and Database Right [June 2018]. Ordnance 866 Survey (Digimap Licence). 867 868 Fig. 3. Modal mineralogy of the soil samples collected throughout Cornwall 869 and Devon, SW England, grouped by lithology according to Table 2. Charts 870 A, C and E show relative abundance of major (>10%) and minor (1-10%) 871 minerals, Charts B, D and F show relative abundance of minor (1-10%) and

872 trace (<1%) minerals. 873 874 Fig. 4. Diagram showing the QEMSCAN false colour particle images for 875 representative samples from the six major bedrock groups in SW England. 876 (A) Sample 10/1 St Austell Granite, (B) Sample 12/1 Nare Head Dolerite, (C) 877 Sample 14/1 Traboe Hornblende Schist (Traboe Cumulate Complex), Lizard 878 Ophiolite Complex, (D) Sample 23/1 Devonian Polzeath Slate Formation 879 metamorphosed mudstone, (E) Sample 9/1 Devonian Staddon Formation 880 metamorphosed sandstone and (F) Sample 30/1 Carboniferous Teign 881 (Newton) Chert Formation. © British Crown Owned Copyright 2018/AWE 27

OFFICIAL

882 883 Fig. 5. Diagram showing the QEMSCAN false colour particle images for two 884 soil samples developed on Devonian metasedimentary bedrock units from SW 885 England. (A) Sample 9/1 Staddon Formation metamorphosed sandstones, 886 (B) Sample 29/1 Tavy Formation metamorphosed mudstone. The quartz grain 887 size and texture is a key discriminator of different sedimentary grades. 888 889 Fig. 6. Modal mineralogy of the 6 soil samples collected from sampling

890 location 24 Porthtowan Formation bedrock, near Allet, Truro, Cornwall, along 891 a 250 m transect. (A) Relative abundance of major (>10%) and minor (1- 892 10%) minerals. (B) Relative abundance of minor (1-10%) and trace (<1%) 893 minerals. 894 895

OFFICIAL

896 897 Figure 1. 898

OFFICIAL

899 900 901 Figure. 2 902

OFFICIAL

903 904 Figure 3

905

OFFICIAL

906 907 908 Figure 4. 909 910

OFFICIAL

911 912 Figure 5 913

914 915 Figure 6 916

917 918

OFFICIAL

919 Table 1. Sampling locations, bedrock geology, OS grid reference and land use.

Rock type Formation Age Land use Grid Reference

1/1 Granite Lands End Granite Permian Resown grassland SW45821 25934

2/1 Dolerite Dolerite Devonian Pasture SW54743 29115

Tregonning- Managed 3/1 Granite Permian SW60211 27618 Godolphin Granite grassland Managed 4/1 Gneiss Kennack Gneiss Devonian (?Famennian) SW72090 16757 grassland Old Lizard Head Managed 5/1 Mica schist ?Cambrian SW80076 22399 Formation grassland Devonian (Lochkovian- 6/1 Meta-mudstone Dartmouth Group Pasture SX14051 53330 Pragian)

7/1 Meta-mudstone Meadfoot Group Devonian (Pragian-Emsian) Crops - maize SX15377 58140

Devonian-Carboniferous 8/1 Meta-mudstone Saltash Formation Resown grassland SX15571 63977 (Emsian-Tournasian)

9/1 Meta-sandstone Staddon Formation Devonian (Emsian) Pasture (very wet) SX13368 63620

10/1 Granite St Austell Granite Permian-Carboniferous Crops - maize SX05528 56036

Meta-sandstone / 11/1 Dodman Formation Early Devonian Pasture SX00095 40220 mudstone Devonian (Givetian- Managed 12/1 Dolerite Nare Head Dolerite SW92165 37879 Frasnian) grassland Roseland Breccia Devonian (Givetian- Managed 13/1 Meta-mudstone SW74187 24719 Formation Frasnian) grassland (wet) Traboe Hornblende Managed 14/1 Schist Schist (Traboe Devonian (?Emsian) SW 74614 21750 grassland Cumulate Complex) Trelan and Crousa Managed 15/1 Gabbro Devonian SW77884 21082 Gabbro grassland Devonian (Emsian) (mantle Managed 16/1 Serpentinite Lizard Serpentinite SW77486 18004 exhumation) grassland Managed 17/1 Granite Carnmenellis Granite Permian SW72821 34663 grassland Mylor Slate Devonian (Frasnian- Managed 18/1 Meta-mudstone SW80249 37757 Formation Famennian) grassland Trevose Slate Devonian (Frasnian- Managed 19/1 Meta-mudstone SX07049 77239 Formation Famennian) grassland

20/1 Granite Bodmin Granite Permian Moorland (grazed) SX10137 78066

Tredorn Slate Managed 21/1 Meta-mudstone Devonian (Famennian) SX07053 85000 Formation grassland Basaltic Tintagel Volcanic 22/1 Carboniferous (Visean ) Coastal grassland SX04879 86451 calcareous tuffs Formation Polzeath Slate Devonian (Frasnian- Managed 23/1 Meta-mudstone SW96717 77740 Formation Famennian) grassland Porthtowan Devonian (Eifelian- Grassland margin 24/1 Meta-mudstone SW78531 47936 Formation Frasnian) to cultivated field Meta-sandstone / Devonian (Givetian- Managed 25/1 Portscatho Formation SW71594 24780 mudstone Frasnian) grassland Managed 26/1 Meta-mudstone Bovisand Formation Devonian (Pragian-Emsian) SX23501 58986 grassland Whitsand Bay Devonian (Lochkovian- Managed 27/1 Meta-mudstone SX33649 54060 Formation Pragian) grassland Devonian (Frasnian- Managed 28/1 Meta-mudstone Torpoint Formation SX39086 57333 Famennian) grassland Devonian (Frasnian- Grazed 29/1 Meta-mudstone Tavy Formation SX40710 63197 Famennian) pastureland Teign (Newton) Chert Grassland margin 30/1 Chert Carboniferous (Visean) SX34764 64914 Formation to cultivated field Managed 31/1 Mudstone Brendon Formation Carboniferous (Visean) SX38587 68631 grassland Milton Abbott Managed 32/1 Basalt-Mudstone Carboniferous (Visean) SX45785 77497 Formation grassland Sandstone- Carboniferous Moorland (near to 33/1 Bealsmill Formation SX50899 81429 Mudstone (Serpukhovian-Bashkirian) Wheal Betsy mine)

34/1 Dolerite Dolertie Devonian-Carboniferous Moorland (grazed) SX52403 80699

35/1 Granite Dartmoor Granite Permian Moorland SX55354 75059

OFFICIAL

920 Table 2. Soil sample numbers, laboratory codes, number of particles 921 measured and number of individual EDS analyses. 922

Sample Block Code Number of particles Number of EDS analyses 1/1 16HG31 5017 170581 2/1 16HG32 5087 340966 3/1 16HG33 5257 95018 4/1 16HG34 5159 183139 5/1 16HG35 5009 179889 6/1 16HG36 5348 144111 7/1 16HG37 5051 487685 8/1 16HG38 5065 344579 9/1 16HG39 5362 60501 10/1 16HG3A 5249 199123 11/1 16HG3B 5152 200560 12/1 16HG3C 5077 214346 13/1 16HG3D 5060 153689 14/1 16HG3E 5570 144240 15/1 16HG3F 5109 121620 16/1 16HG3G 5170 98969 17/1 16HG3H 5148 72579 18/1 16HG3I 5145 109501 19/1 16HG3J 5210 89146 20/1 16HG3K 5392 73202 21/1 16HG3L 5280 60674 22/1 16HG3M 5257 73762 23/1 16HG3N 5080 325097 24/1 16HG3O 5385 113463 25/1 16HG3P 5304 117203 26/1 16HG3Q 5021 102166 27/1 16HG3R 5105 112770 28/1 16HG3S 5191 172621 29/1 16HG3T 5197 135687 30/1 16HG3U 5087 197134 31/1 16HG3V 5645 62752 32/1 16HG3W 5524 110229 33/1 16HG3X 5075 50734 34/1 16HG3Y 5098 81299 35/1 16HG3Z 5220 77215 24/1 16HG3AA 5213 95367 24/2 16HG3AB 5057 67539 24/3 16HG3AC 5066 87886 24/4 16HG3AD 5075 119598 24/5 16HG3AE 5123 107509 923 © British Crown Owned Copyright 2018/AWE 35

OFFICIAL

924 Table 3. Mineral groupings used to process automated mineralogy data. 925

Mineral Grouping Description Quartz Quartz. May include other silica minerals Plagioclase feldspar Plagioclase (albite to anorthite solid solution) K feldspar K feldspar such as orthoclase and microcline Muscovite. May include alteration products after feldspars such as Muscovite sericite. Illite also reports to this category. Biotite Biotite and phlogopite. Would also include glauconite Chlorite Chlorite. Man-made slags would also report to this category Kaolinite Kaolinite, halloysite. dickite Fe silicates Nontronite and other Fe silicates Mg silicates such as serpentine group minerals, orthopyroxene, Mg silicates olivine Hornblende Hornblende and other amphiboles Tourmaline Tourmaline group minerals Epidote Group Epidote group minerals Sillimanite Sillimanite, andalusite, kyanite Topaz Topaz Other silicates Garnet. Al silicates Calcite Calcite, ankerite, dolomite, magnesite Rutile Rutile, titanite (sphene) Ilmenite Ilmenite Fe-Mn Ox Fe oxides (magnetite, hematite), Mn oxides Chromite Chrome spinel Apatite Apatite, biogenic apatite Zircon Zircon Xenotime Xenotime Ce phosphate Monazite Fe / Cu sulphides Pyrite, chalcopyrite Gypsum Gypsum Others Any other mineral not reported above 926 927 928 929 930 931 932 933 934 935 936

OFFICIAL

937 Table 4. Modal mineralogical data for the 35 soil samples analysed throughout 938 Cornwall and Devon, SW England. 939

1-1 2-1 3-1 4-1 5-1 6-1 Quartz 47.17 18.57 52.89 21.80 24.17 29.10 Plagioclase feldspar 9.38 9.06 10.39 17.45 13.24 4.80 K-Feldspar 19.18 7.36 16.95 13.71 10.05 26.77 Muscovite 2.44 1.11 3.82 0.90 2.26 8.21 Biotite 7.01 16.26 5.03 4.26 12.40 15.93 Chlorite 11.83 35.42 7.74 23.50 22.30 11.51 Kaolinite 0.84 0.29 1.71 0.08 3.18 1.30 Fe silicates 0.44 0.80 0.27 1.63 0.61 0.67 Mg Silicates 0.02 0.12 0.00 11.05 0.11 0.01 Hornblende 0.19 2.56 0.01 4.31 2.62 0.02 Tourmaline 0.06 0.90 0.03 0.15 0.48 0.01 Epidote Group 0.51 1.52 0.65 0.02 5.97 0.34 Sillimanite 0.00 0.00 0.01 0.00 0.00 0.00 Topaz 0.01 0.00 0.01 0.00 0.00 0.00 Other silicates 0.15 1.86 0.07 0.31 1.39 0.25 Calcite 0.10 0.32 0.08 0.08 0.11 0.25 Rutile 0.45 1.77 0.19 0.26 0.70 0.52 Ilmenite 0.08 0.70 0.04 0.02 0.18 0.02 Fe-Mn Ox 0.08 0.60 0.01 0.32 0.15 0.13 Chromite 0.01 0.02 0.01 0.07 0.01 0.04 Apatite 0.00 0.69 0.01 0.06 0.01 0.04 Zircon 0.02 0.02 0.01 0.01 0.01 0.03 Xenotime 0.01 0.01 0.02 0.00 0.01 0.00 Ce Phosphate 0.00 0.00 0.00 0.00 0.00 0.00 Fe/Cu sulphides 0.01 0.00 0.00 0.00 0.00 0.01 Gypsum 0.00 0.01 0.00 0.00 0.00 0.00 Others 0.01 0.02 0.01 0.00 0.02 0.00 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954

OFFICIAL

955 Table 5. Key mineralogical parameters for the six granite derived soils compared 956 with the classification of the exposed granite types from (Simons et al., 2016). 957

1 - Lands End 3 - Tregonning 10 - St Austell 17 - Carnmenellis 20 - Bodmin 35 - Dartmoor Granite Godolphin Granite Granite Moor Granite Granite Granite Kfsp 19.18 16.95 18.62 22.39 29.09 39.31 Plag 9.38 10.39 8.89 8.79 19.33 11.11 Biot 7.01 5.03 6.22 4.21 1.30 4.88 Musc 2.44 3.82 7.87 4.47 4.47 3.70 Kaol 0.84 1.71 3.93 1.79 1.30 1.67 Tourm 0.51 0.65 2.40 0.52 1.00 1.58 Sillim 0 0.01 0.01 0.03 0.23 0 Topaz 0.01 0.01 0 0.02 0.21 0 Granite G3 biotite G1 two mica G3 biotite G1 two mica G1 two mica G3 biotite types granite granite granite granite granite granite G4 G5 topaz G4 G2 Muscovite G2 G4 tourmaline granite tourmaline granite Muscovite tourmaline granite granite granite granite 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987

OFFICIAL

988 Table 6. Modal mineralogical data for the soil samples collected along a 250 m 989 transect at Location 24, near Allet, Truro, Cornwall, with replicate sample 24/1. 990

24-1(a) 0 m 24-1(b) 0 m 24-2 10 m 24-3 50 m 24-4 100 m 24-5 250 m Quartz 37.69 40.77 32.84 31.91 32.20 36.42 Plagioclase feldspar 3.06 3.41 3.48 2.41 3.68 3.91 K-Feldspar 17.22 16.60 17.68 17.51 22.13 20.75 Muscovite 13.65 14.45 20.03 20.31 13.37 11.17 Biotite 17.49 15.12 16.71 17.05 17.67 18.66 Chlorite 7.00 6.58 6.11 7.38 6.98 6.23 Kaolinite 1.40 1.15 1.70 1.66 1.54 1.17 Fe silicates 0.25 0.26 0.18 0.31 0.22 0.21 Mg Silicates 0.01 0.00 0.05 0.00 0.01 0.03 Hornblende 0.01 0.00 0.01 0.00 0.03 0.05 Tourmaline 0.97 0.85 0.46 0.79 0.45 0.47 Epidote Group 0.02 0.01 0.00 0.00 0.03 0.02 Sillimanite 0.00 0.00 0.01 0.00 0.00 0.00 Topaz 0.00 0.00 0.02 0.00 0.00 0.00 Other silicates 0.20 0.28 0.10 0.12 0.13 0.26 Calcite 0.52 0.07 0.08 0.04 1.04 0.09 Rutile 0.29 0.30 0.40 0.42 0.39 0.39 Ilmenite 0.01 0.03 0.01 0.01 0.01 0.01 Fe-Mn Ox 0.03 0.03 0.02 0.01 0.03 0.09 Chromite 0.01 0.01 0.00 0.00 0.00 0.00 Apatite 0.15 0.04 0.00 0.00 0.00 0.00 Zircon 0.01 0.01 0.04 0.04 0.02 0.02 Xenotime 0.00 0.01 0.01 0.01 0.01 0.00 Ce Phosphate 0.01 0.01 0.01 0.01 0.01 0.01 Fe/Cu sulphides 0.00 0.00 0.00 0.00 0.03 0.00 Gypsum 0.01 0.00 0.00 0.00 0.01 0.01 Others 0.02 0.01 0.03 0.00 0.02 0.03 991 992 993